This paper is mainly concerned with the problem of distributing a data base (i.e., a set of segments) in a computer network system so as to facilitate parallel searching. In our distributed data base model, we assume that all segments are stored in nodes. Each time a query occurs, all nodes are searched concurrently. For convenience, we define the time required to access a segment from any node as a time unit. For a network withdnodes, the response time of a query is then identical to the maximum (n1, n2, …,nd), whereni, is the number of segments that satisfies the query and is stored in nodei. Unfortunately, the solution for finding an optimal way to organize a distributed data base for parallel searching is still at large. In other words, given a data base, there is no efficient polynomial time algorithm for finding an optimal arrangement of segments onto nodes. In this article, we shall present a “heuristic algorithm” based upon a multivariant analysis method in statistics to distribute a data base in a network system. Some experimental results will show that our method is indeed feasible and effective.